(In reply to Hillwood Yang from comment #5) > (In reply to Fabian Vogt from comment #4) > > IMO, "resources_big" should be avoided anyway. It leads to longer compile > > times and bigger executables, and the way it works is pretty hacky and > > fragile. > > Especially when the resource data is that big that compiling runs out of > > memory, it's much better to just read the data files directly instead. > > > > In this particular case it appears to affect test cases, which aren't > > shipped in packages anyway. So there's no reason to use resources for those. > > I can not avoid "resources_big" on Tumbleweed. It leads to be the hign > memory usage issue without "resources_big", it needs 14GB memory to complete > the building for this project > https://build.opensuse.org/package/show/X11:Deepin:Factory/deepin-reader. > But this serious memory issue does not occur on Leap. The data should not be embedded as resources but read from the filesystem directly. https://github.com/linuxdeepin/deepin-reader/blob/dev/tests/ut_defines.h should use QFINDTESTDATA instead of resources. Or just don't build the tests at all, the .spec file doesn't run them anyway.